How to exclude the unmapped pair reads in bam files

[vincent lee]

There is a bam file with the pair-end reads alignment.

I want to exclude the reads only when neither of pair reads are mapped onto the genome.
What command I should type？

I have tried samtools view -f 2, which just removed the pair if either one is unmapped.( The same as samtools view -F 12 )

Besides, samtools view -F 4 or samtools view -F 8 will remove a read while the other is mapped.

Thanks!

Best regards！

[Lluc]

I would suggest to go through the file a couple of times, first to take the half mapped reads, and then the ones completely mapped:

samtools view -F 4 -f 8 input.bam > only_this_end_mapped.sam
samtools view -F 8 -f 4 input.bam > only_that_end_mapped.sam
samtools view -F 12 input.bam > both_ends_mapped.sam

And then, merge them.

Alternatively, you could select the reads that you want to exclude, write their names in a file, and use that list as a filter with Picard's FilterSamReads:

samtools view -f 12 input.bam | cut -f 1 > blacklist
java -jar FilterSamReads.jar INPUT=input.bam FILTER=excludeReadList READ_LIST_FILE=blacklist OUTPUT=clean.bam

[vincent lee]

Thanks for your reply Lluc!

It seem no other method can be faster... I would like to try the first one you mention above.
Hoping someone could write a program in C or C++, which can make it faster.

Best regards!